WSU college studies A.I. in farming

PULLMAN - Washington State University's College of Agricultural, Human, and Natural Resource Science received a $20 million federal grant in 2021 from USDA to develop artificial intelligence (AI) tools and approaches to support real-time site-specific decision-making at the farm scale.

Associate Dean and Director, ARC, Scot Hulbert, confirmed that the project is a collaboration between AgAID Institute and the National Science Foundation.

Lead Institute Professor Ananth Kalyanaraman explained that what they are trying to do in their research is to see how to harness the power of data and scientific understanding.

Kalyanaraman explained that there are three ways that AI can help farmers with making decisions on the farm. First, the AI models can predict incoming or temperature changes.

He noted that it would also help the farmer know what he needs to use for irrigation.

"When to spray the crops," he said, "There's a lot of daily decisions that the farmers have to make."

Associate Professor of Precision Ag Interim Director Lav Khot explained that most of the data they've already fed into the AI models come from the farms.

He explained that they use the data to see which model works in agriculture.

This data then goes to the researchers to make the models.

"We are trying to collect to understand the heat source, feed that to land and irrigation earlier," Khot said.

Khot and Kalyanaraman noted that Washington grows many high-value crops, such as apples, grapes, and blueberries.

Kalyanaraman explained that these are sitting on land that requires water. "Water is an important resource," he said.

Khot added that irrigation plays a significant role in this. In addition, the team has been collecting soil and canopy data with Dr. Marcus Taylor to improve the quality of the berries.

He explained that Taylor has a lot of data on that aspect of the research.

Kalyanaraman stated that they have models based on monitoring temperature.

Khot added that this data could help fruit be harvested at the correct time so it does not have sunburn.

"We are feeding that to the AI team creating the models that can better help us predict," he said.

He noted that if the temperature drops, the fruit may become damaged.

Khot stated that weather is an essential aspect of their research and that the National Weather Service is less accurate in the micro-content.

"AgAID is trying to improve the quality of the weather data," Khot said.

Hulbert explained that a lot of robotics research is going on in the orchards, from harvest aides to actual harvesters, pruners, leaf thinning, and blossom thinning.

"They're looking at robotics to stop as much labor as we can save on cost," he said, adding that much of the AI project is about labor, including optimization and reducing labor costs.

Kalyanaraman stated that one of their key questions is how AI can partner with people in the fields.

There would be models of robots that could prune and harvest most of these crops, like apples and so on.

"It's difficult to find high-skilled, high-quality labor," he said, noting that there are a lot of uncertainties that farms have to deal with that create a lot of loss.

The other possibility Kalyanaraman mentioned is the human doing the tasks, and a smart held device gives some guidance.

"That could be a powerful tool in a couple of ways," he said, adding that it would give consistency to workflow.

"It can train low-skill workers to bridge the gap between low and high-skilled workers," he said.

He stated that tools and techniques come in different forms and that farmers already use some support systems.

Hulbert explained that biosensors collect information throughout the growing season, and smart orchards collect many different types of sensors.

These sensors include soil sensors, which tell the moisture in a particular sport or the fertility in the soil.

There are also images taken that can predict fruit load and programs optimized to tell where to make cuts or what branches to cut off.

Other sensors include ones that can tell what the temperature is on the fruit once the sun is shining on them to see if it needs a bit of mist sprayed on it.

"It's really about taking all that data and then optimizing models to deliver inputs like water and nutrients and stuff at the optimal amounts and times throughout the year," Hulbert said.

Kalyanaraman explained that site-specific data is useful and can now be put on webpages, and apps are in development now.

"The interface should be lightweight to use," he said, adding that it would engage with the farm.

He explained that users of the interfaces could put the data in the cloud and see it on their mobile phones if they wanted to do that.

Khot stated that smartphone data is used, and they are getting quality data through their research and collecting more data that the AI team will need.

He further stated that they now have a grape model and are hosting that on their website. This season they plant to test that on some ag growers to decide what mitigation they need to do.